Content
@
https://opensea.io/collection/dev-21
0 reply
0 recast
2 reactions
Nastya
@nastya
Been diving into EigenLayer AVS lately, sharing my super general and practical understanding of what is it for: * You have some task – like a piece of program, request, onchain action etc. It produces an output. * You need to prove to a skeptical 2nd party that it’s correct and truly came from that task. * Here comes a network of machines (AVS), that can re‑execute the task, confirm the result, and get paid for it. * The 2nd party can cryptographically verify that machines really did this job. * But what if some of these machines were malicious? * That's staking comes in – machines put their money to join the network, earn rewards, but risk penalties for bad behavior. EigenLayer essentially creates a marketplace for computational trust. But another practical question that comes up to me: what are the actual things you’d want to verify with this?
5 replies
5 recasts
55 reactions
Monteluna
@monteluna
Verification of LLM output. Airport checkin. Local in person geotagging. Also that system a while ago that allowed you to do zkttls swaps between crypto and venmo.
1 reply
0 recast
2 reactions
Nastya
@nastya
Oh, I’ve been looking into AI/LLM use cases, and it seems like there are a lot of nuances. To verify LLM outputs, the process would need to be deterministic (same input gives the same output), but that kind of contradicts how LLMs are typically used. There’s this article from a company that run AVSs on the topic, but by “AI determinism” they mean different things - like hardware consistency, model parameters, consensus etc https://www.layer.xyz/news-and-insights/deterministic-ai
1 reply
0 recast
2 reactions
